Figure 6

The fitting procedure involves 2500 local optimisations for each patient. The simulation of the model at each optimisation step requires one to track the zero-crossing phase in order to provide stimulation at the right phase. The phase-tracking ability of the scheme is satisfactory when compared to the actual Hilbert phase (left, detailed in Supplementary Fig. 2 in Appendix I). The optimiser minimises a cost function that includes the comparison of three tremor dynamics features (tremor PSD, tremor envelope PSD, tremor envelope PDF) plus the bPRC against the data (middle). Response curves are obtained the same way for the data and the model. Following a second optimisation of the 20 best results with a finer time step, a best set of parameters comes out of the procedure, and the model bARC can be compared against the data bARC. More details on the fitting procedure are given in Appendix F